Overall I would agree with you, but I start feeling that this „iron law“ isn’t as simple as that. After all, humans have limited „context window“ too — we don’t remember every small detail on a large project we have been working on for several years. Loose coupling and modularity helps us and can help LLM to make the size of the task manageable if you don’t ask it to rebuild the whole thing. It’s not the size that makes LLMs fail, but something else, probably the same things where we may fail.

▲

timr 4 days ago | parent [-]

Humans have a limited short-term memory. Humans do not literally forget everything they've ever learned after each Q&A cycle.

(Though now that I think of it, I might start interrupting people with “SUMMARIZING CONVERSATION HISTORY!” whenever they begin to bore me. Then I can change the subject.)

▲

ivan_gammel 4 days ago | parent | next [-]

LLMs do not „forget“ everything completely either. Probably all major tools by now consume information from some form of memory (system prompt, Claude.md, project files etc) before your prompt. Claude Code rewrites the Claude.md, ChatGPT may modify the chat memory if it finds it necessary etc.

▲

timr 4 days ago | parent [-]

Writing stuff in a file is not “memory” (particularly if I have to do it), and in any case, it consumes context. Overrun the context window, and the tool doesn’t know about what is lost.

There are various hacks these tools take to cram more crap into a fixed-size bucket, but it’s still fundamentally different than how a person thinks.

	▲	ivan_gammel 4 days ago \| parent [-]
		> Writing stuff in a file is not “memory” Do you understand yourself what you just said? File is a way to organize data in memory of a computer by definition. When you write instructions to LLM, they persistently modify your prompts making LLM „remember“ certain stuff like coding conventions or explanations of your architectural choices. > particularly if I have to do it You have to communicate with LLM about the code. You either do it persistently (must remember) or contextually (should know only in context of a current session). So word „particularly“ is out of place here. You choose one way or another instead of bring able to just tell that some information is important or unimportant long-term. This communication would happen with humans too. LLMs have different interface for it, more explicit (giving the perception of more effort, when it is in fact the same; and let’s not forget that LLM is able to decide itself on whether to remember something or not). > and in any case, it consumes context So what? Generalization is an effective way to compress information. Because of it persistent instructions consume only a tiny fraction of context, but they reduce the need for LLM to go into full analysis of your code. > but it’s still fundamentally different than how a person thinks. Again, so what? Nobody can keep in short-term memory the entire code base. It should not be the expectation to have this ability neither it should not be considered a major disadvantage not to have it. Yes, we use our „context windows“ differently in a thinking process. What matters is what information we pack there and what we make of it.

▲

faangguyindia 3 days ago | parent | prev | next [-]

the "context" is the short term memory equivalent of LLM.

Long term memory is its training data.

▲

BeetleB 4 days ago | parent | prev [-]

Both true and irrelevant.

I've yet had the "forgets everything" to be a limiting factor. In fact, when using Aider, I aggressively ensure it forgets everything several times per session.

To me, it's a feature, not a drawback.

I've certainly had coworkers who I've had to tell "Look, will you forget about X? That use case, while it look similar, is actually quite different in assumptions, etc. Stop invoking your experiences there!"